Without this patch, the accompanying selftest never wakes up from HLT
in L2. If you can get the selftest to work without this patch, feel
free to drop it.
On Mon, May 24, 2021 at 8:43 AM Paolo Bonzini <pbonzini@xxxxxxxxxx> wrote:
>
> On 21/05/21 01:03, Jim Mattson wrote:
> > At present, there are two reasons why kvm_check_nested_events may
> > return a non-zero value:
> >
> > 1) we just emulated a shutdown VM-exit from L2 to L1.
> > 2) we need to perform an immediate VM-exit from vmcs02.
> >
> > In either case, transition the vCPU to "running."
> >
> > Signed-off-by: Jim Mattson <jmattson@xxxxxxxxxx>
> > Reviewed-by: Oliver Upton <oupton@xxxxxxxxxx>
> > ---
> > arch/x86/kvm/x86.c | 4 ++--
> > 1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> > index d517460db413..d3fea8ea3628 100644
> > --- a/arch/x86/kvm/x86.c
> > +++ b/arch/x86/kvm/x86.c
> > @@ -9468,8 +9468,8 @@ static inline int vcpu_block(struct kvm *kvm, struct kvm_vcpu *vcpu)
> >
> > static inline bool kvm_vcpu_running(struct kvm_vcpu *vcpu)
> > {
> > - if (is_guest_mode(vcpu))
> > - kvm_check_nested_events(vcpu);
> > + if (is_guest_mode(vcpu) && kvm_check_nested_events(vcpu))
> > + return true;
>
> That doesn't make the vCPU running. You still need to go through
> vcpu_block, which would properly update the vCPU's mp_state.
>
> What is this patch fixing?
>
> Paolo
>
> > return (vcpu->arch.mp_state == KVM_MP_STATE_RUNNABLE &&
> > !vcpu->arch.apf.halted);
> >
>
